- Home
- Skills
- Code Quality & Review
- ApprovalLens Companion — Agent Action Review
ApprovalLens Companion — Agent Action Review
Compare an operation with the original task and explicit file permissions, identify extra effects, and get a structured recommendation with its uncertainties visible.
Free
Works with the AI Agents you already use
ApprovalLens Companion — Agent Action Review
Example session with this skill installed
Use ApprovalLens Companion to review this proposed action without executing it.
Task: update the greeting in demo.txt; preserve all other files.
Rules: read_file and write_file grants for demo.txt only.
Agent explanation: “Update the greeting.”
Proposed operations: read demo.txt; write “hello ApprovalLens” to demo.txt; delete notes.txt.
Give me the effects, rule conflicts, uncertainties and the smallest next step.
- Read your context and instructions
- Compiled the approvallens companion
Assessment: POLICY CONFLICT
The proposed deletion of notes.txt is outside the supplied permissions and contradicts the instruction to preserve other files. This is an advisory analysis of the supplied example, not a live engine result.
Intent
Update the greeting in demo.txt while preserving all other files. The supplied policy grants only reading and writing demo.txt.
Effects
| Operation | Target | Evidence | Rule fit |
|---|---|---|---|
| Read | demo.txt | Explicit proposed read | Matches read_file grant |
| Write | demo.txt | Proposed replacement: hello ApprovalLens | Matches write_file grant and task intent |
| Delete | notes.txt | Explicit proposed deletion | No delete grant; conflicts with preservation instruction |
Explanation mismatch
“Update the greeting” describes the demo.txt change but omits the unrelated deletion. The explanation is incomplete for the full proposed action.
Unknowns
No filesystem or live extension event was inspected. Existing file contents, actual host tool behavior and execution outcome are not verified. The stated conflict follows directly from the supplied operations and grants; those unknowns do not remove it.
Next step
Remove the notes.txt deletion from the proposal. Review the narrowed demo.txt operation separately in your host. Do not broaden permissions merely to make this request pass. Nothing was executed, blocked by a runtime, or authorized by this advisory review.
Connects securely to your tools. The creator never sees your data.
About this skill
Understand the action before granting permission
ApprovalLens Companion helps developers review what a coding agent proposes to do—not just the explanation it gives. Compare an operation with the original task and explicit file permissions, identify extra effects, and get a structured recommendation with its uncertainties visible.
What you get
- An advisory review: task intent, effects table, explanation mismatches, unknowns and next steps.
- Narrow ApprovalLens policy drafts with separate, exact-path read/write grants. Existing restrictions are preserved.
- Help interpreting local findings, hosted semantic findings, human decisions and reported outcomes without confusing them.
- A reference for the separate Cursor extension setup and official resources.
How to use it
Provide the original task, proposed tool call or command, the agent’s explanation and the applicable rules. Ask for an advisory review before executing the action. For policy drafting, describe a concrete task and the files it needs. Review the draft yourself before applying any permission change.
Clear boundaries
This is a Markdown skill, not a security boundary. It does not intercept or block tools, execute the reviewed operation, approve requests, or activate rules. Its analysis depends on the evidence supplied and your agent’s reasoning. Unknown script behavior remains unknown.
Native enforcement is available through the separate ApprovalLens Cursor extension for supported hook-delivered operations in one local macOS/Linux workspace. Hosted Nemotron review requires a separately supplied private access token; this listing includes no inference credits or hosted access.
Package
SKILL.md, one policy/setup reference and the MIT license. No executable scripts, bundled credentials or required external service for reviewing supplied examples. Free companion to the open-source ApprovalLens project.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.
Advisory reasoning only; not a security boundary or execution authorization. Does not intercept, block or execute native tool calls. Cannot prove arbitrary shell/script behavior from incomplete evidence. Native enforcement requires the separate Cursor extension and supported hooks; coverage is not universal. No hosted token or inference allowance included.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 19 days ago
- Free with an Agensi account
Permissions
Read-only access to user-authorized task files and existing .approvallens/rules.json when needed. Supplied examples can be reviewed without filesystem access. No shell execution, file writes, environment-variable access or external API calls are required by this skill. Policy drafts are shown for human review; they are not activated.